AI Native Observability
atrim.ai is an AI-native observability platform that helps engineering teams understand and debug distributed systems in seconds, not hours. Point your existing OpenTelemetry at our endpoint — no agent, no operator, nothing to install in your cluster — and analysis starts on its own. Unlike traditional tools that hand you dashboards and leave you to find the problem, atrim runs continuous statistical analysis across your telemetry — baselining every service, ranking the critical paths, and scoring blast radius — then enriches each finding with LLM evaluation that explains, in plain language, what changed and what it affects. atrim reads your telemetry and tells you what's wrong — it never acts on your systems and holds none of your credentials.
Problems We Solve
- Blank-dashboard paralysis — Traditional tools hand you dashboards and leave you to find the problem. atrim ranks the locations where latency and errors accumulate and puts the path that matters first.
- Alerts without context — Signals are scored against each service's own baseline history, so a real shift stands out instead of a static threshold firing on noise.
- No time to build observability — No dashboards to build, no thresholds to tune, nothing to install. Analysis starts on its own the moment telemetry arrives.
- The long tail that never justified the cost — Apps that can't yet justify a tier-1 observability bill still get first-class analysis — so they're already covered if they take off.
Who It's For
- Teams that own a handful of services with no SRE assigned
- Owners of the platform long tail — apps that never justified a tier-1 observability bill
- Teams shipping AI-built apps that now carry real users
Team
We spent our careers building the tier-1 observability stack — Fortune 500 APM, carrier-scale service assurance, AIOps pipelines. Atrim is what we're building for all the new AI-built apps those tools were never priced or designed for. It's a different audience, too: not a NOC, but teams that have moved from DevOps, to SRE, to AI-assisted ops.
Founded StreamWeaver, an AIOps data integration platform acquired by BMC Software in 2021, where he led product development and launched a new OpenTelemetry product. Previously founded J9 Technologies, a leading APM consulting firm serving Fortune 500 companies. Early career includes The Middleware Company (TheServerSide.com), Momentum Software, and Mercury Interactive.
Has spent 25+ years building the network and service management systems that enterprise operations teams run on. Most recently VP of Software Engineering at StreamWeaver, the AIOps data integration platform acquired by BMC Software in 2021. Before that, Principal Architect and development manager for network performance management at Persistent Systems, and a Senior Technical Staff Member in IBM's Tivoli CTO office. Earlier he was Principal Architect at Micromuse — whose Netcool service assurance suite IBM acquired for $865M in 2006 — and CTO & Chief Architect at Quallaby, whose PROVISO carrier-scale performance platform Micromuse acquired in 2005 and shipped as Netcool/Proviso.
Key Features
Ranked, explained findings
Critical Paths and Metric Signals rank where to look using statistics; per-task routing across Kimi, Claude, Qwen, and Inkling models evaluates each finding and explains it in plain language
Service graph from your telemetry
Every service and call drawn from your traffic — with rate, errors, and latency on nodes and edges, and blast radius up and downstream
Ask in plain English
Describe what you want in the Traces, Logs, and Metrics explorers; atrim compiles it to a structured query over your raw OpenTelemetry — sub-second, powered by ClickHouse
Incidents that track the condition
A condition opens an incident and notifies you by email or webhook; when the condition clears, the incident resolves — finding and resolution on one record